NAREOR: The Narrative Reordering Problem

Many implicit inferences exist in text depending on how it is structured that can critically impact the text's interpretation and meaning. One such structural aspect present with chronology order of its presentation. For narratives or stories, this known as narrative order. Reordering a temporal, causal, event-based, other readers draw from it, which turn have strong effects both interestingness. In paper, we propose investigate task Narrative (NAREOR) involves rewriting given story different while preserving plot. We dataset, NAREORC, human rewritings stories within ROCStories non-linear orders, conduct detailed analysis it. Further, novel task-specific training methods suitable evaluation metrics. perform experiments NAREORC using state-of-the-art models BART T5 extensive automatic evaluations. demonstrate although our decently, NAREOR challenging potential for further exploration. also two applications NAREOR: generation more interesting variations serving adversarial sets temporal/event-related tasks, besides discussing prospective ones, pedagogical setups related to language skills like essay writing medicine involving clinical narratives.